🤖 AI Agents Weekly: GLM-5.2, Claude Code Artifacts, Qwen-Robot Suite, Codex Skills, Block's Builderbot, SpatialClaw, and More
GLM-5.2, Claude Code Artifacts, Qwen-Robot Suite, Codex Skills, Block's Builderbot, SpatialClaw, and More
In today's issue:
Z.ai open-sources frontier GLM-5.2
Claude Code ships interactive Artifacts
Qwen launches the Robot Suite
Codex turns demos into skills
Block's Builderbot writes 15% of code
Flue 1.0 reimagines the agent harness
Vercel debuts the eve framework
Cursor launches Origin code hosting
Perplexity adds Brain memory to Computer
OpenRouter ships the Fusion API
NVIDIA's SpatialClaw codes spatial reasoning
ENPIRE self-improves robot policies
SkillsBench 1.1 audits agent skills
DeepMind maps an AI Control Roadmap
OpenAI trains broadly beneficial models
Anthropic measures returns to expertise
And all the top AI dev news, papers, and tools.
Top Stories
Z.ai Open-Sources GLM-5.2, a Frontier Model with a 1M-Token Context
Z.ai released GLM-5.2, an open-weights model built for long-horizon coding and agentic work, with a usable 1-million-token context window and selectable reasoning effort.
Agentic focus: Significant gains on coding and agentic tasks, tuned for large-scale implementation, automated research, performance optimization, and complex debugging.
Two effort levels: GLM-5.2 (max) pushes peak performance while GLM-5.2 (high) balances quality against token efficiency.
Long context: A 1M-token window with up to 128K output tokens, sized to hold full-repository state, API contracts, and prior decisions across long sessions.
Open and compatible: MIT-licensed open weights, working out of the box with Claude Code, Cline, OpenCode, Roo Code, Goose, and Crush.
Omnigent Is an Open-Source Meta-Harness for All Your AI Agents
Omnigent is an open-source framework and meta-harness that runs Claude Code, Codex, Cursor, Pi, and custom agents under one interface, with no vendor lock-in. The Apache 2.0 project has crossed 4.2k GitHub stars and targets builders who juggle several coding agents.
Multi-agent supervision: Orchestrate and delegate across many coding agents in one shared session, with real-time team collaboration and session sharing.
Model flexibility: Bring your own API keys, subscriptions, gateways, or Databricks, and switch models without rewrites.
Run anywhere: Sync sessions across terminal, browser, and mobile, with cloud sandbox execution on Modal, Daytona, and Islo.
Governance built in: Policy-based controls for spend caps, approval gates, and tool restrictions, plus custom agents defined in YAML.